Rate Based Protocols Evaluation

This section presents the throughput, end-to-end delay and number of received packets results of rate based congestion techniques compared to TCP. The rate based congestion control mechanisms used are XCP, XCP-b and RCP. These mechanisms use network in- teraction for rate adaptation and congestion control.

Figure 3.12, Figure 3.13 and Figure 3.14 show the obtained results for the 4x4 grid mesh nodes scenario. Figure 3.12 shows the larger throughput of XCP-b. It must be noticed, however, that as XCP-b uses an analytical model that relies on the maximum buffer size of nodes and in complex heuristics, with the increase in the number of mobile nodes, it becomes less accurate and efficient, making XCP-b to obtain results that are very similar to TCP. From the obtained results, it is also possible to see that TCP has a very regular and fair behavior, while RCP and XCP are less efficient, less fair and sometimes show a very erratic and irregular behavior. Although while regarding throughput, the conclusion of more efficiency is not very clear from the delay and number of received packets, TCP has a very clear and improved performance when compared with both XCP and RCP. As previously mentioned, it is possible to outperform fairness with throughput and bandwidth allocation. Thus, from the combination of Figure 3.12 and Figure 3.13, it is possible to conclude that TCP is more fair than XCP and RCP. Comparing to XCP-b, and as more nodes exist in the network, TCP is using more efficiently the medium. It is evident that the use of an analytical model that relies in buffer size is making XCP-b to use the medium inefficiently.

From the obtained results, it is evident that both XCP and RCP do not perform well in a wireless mesh network. XCP and RCP need, to operate, that all nodes in the network exchange information, and they also need to correctly infer the link capacity and available bandwidth. Since they are not inferring correctly those parameters, their performance is

0 100 200 300 400 500 600 700 3 3.5 4 4.5 5 5.5 6 6.5 7 Throughput(Kbps)

Number of Mobile Nodes TCP Avg. Throughput

XCP Avg. Throughput XCP-b Avg. ThroughputRCP Avg. Throughput

Figure 3.12: Rate Based Protocols Average Throughput - 16 Mesh Nodes, Variable Number of Mobile Nodes. 10 100 1000 10000 100000 3 3.5 4 4.5 5 5.5 6 6.5 7 Delay (ms)

Number of Mobile Nodes TCP Avg. Delay

XCP Avg. Delay XCP-b Avg. DelayRCP Avg. Delay

Figure 3.13: Rate Based Protocols Average Average Delay - 16 Mesh Nodes, Variable Number of Mobile Nodes.

significantly reduced, increasing the number of collisions and delay, and obtaining lower throughput values. Figure 3.14 shows that TCP has very good results in terms of received packets, thus, having fewer packet losses. Since XCP-b uses node buffer size, when the network is fully utilized, it is not accurately obtaining available bandwidth values resulting in higher packet losses. The results show that XCP and RCP are less efficient and fair than TCP. While XCP-b has the best results, when mobility and number of nodes increases, its results are comparable to the ones obtained by TCP.

3.4 Performance Evaluation Results 0 2000 4000 6000 8000 10000 12000 14000 3 3.5 4 4.5 5 5.5 6 6.5 7 Number of Packets

Number of Mobile Nodes TCP Avg. Recv. Packets

XCP Avg. Recv. Packets XCP-b Avg. Recv. PacketsRCP Avg. Recv. Packets

Figure 3.14: Rate Based Protocols Average Average Received Packets - 16 Mesh Nodes, Variable Number of Mobile Nodes.

Figure 3.15, Figure 3.16 and Figure 3.17 show the evaluation parameters results for mesh scenarios with a fixed number of 7 mobile nodes and a variable number of mesh nodes (5, 9, 12 and 15 mesh node). The results show that XCP-b obtains the best throughput results. However, as the number of mesh nodes increases XCP-b results are very similar to TCP results. Moreover, XCP-b obtains worse results than TCP for the number of received packets and for delay when the number of mesh nodes is 16. With more routing messages crossing the network, XCP-b is not able to correctly distinguish between transmission traffic from routing traffic, and considering also both routing and communication retransmissions, it becomes less efficient and with higher number of losses.

TCP, once more, presents a very regular and stable behavior. RCP and XCP behav- iors are sometimes very erratic and irregular with different values in terms of throughput, delay and received packets. As XCP and RCP need, to operate, that all nodes in the network exchange information, the number of collisions increases, leading to higher losses, and consequently lower number of packets received and lower throughput and high delay. Thus, XCP and RCP are not allowing nodes to correctly manage their queues, nor ob- taining correctly available bandwidth and link capacity values. XCP and RCP rate based control mechanisms clearly suffer from incorrect control parameters estimation in wireless environments, due specially to the underlying shared medium.

The ad-hoc scenario results are presented in Figure 3.18, Figure 3.19 and Figure 3.20. In Figure 3.18, it is shown that in this particular simulation, the throughput values are very similar for all the evaluated proposals, obtaining XCP-b the best results. It must be noticed that, regarding delay and the number of received packets, it is not so evident that XCP-b outperforms the other evaluated protocols. The best results of XCP-b are, once more, when the network is not heavily utilized, where the XCP-b model is able to

20 40 60 80 100 120 140 160 180 200 4 6 8 10 12 14 16 Throughput(Kbps)

Number of Mesh Nodes TCP Avg. Throughput

XCP Avg. Throughput XCP-b Avg. ThroughputRCP Avg. Throughput

Figure 3.15: Rate Based Protocols Average Throughput - Variable Number of Mesh Nodes, 7 Mobile Nodes. 100 1000 10000 4 6 8 10 12 14 16 Delay(ms)

Number of Mesh Nodes TCP Avg. Delay

XCP Avg. Delay XCP-b Avg. DelayRCP Avg. Delay

Figure 3.16: Rate Based Protocols Average Average Delay - Variable Number of Mesh Nodes, 7 Mobile Nodes.

correctly obtain the available bandwidth and consequently the rate, allowing for a more effective congestion control. However, as the network load increases, XCP-b becomes more inefficient not using properly the medium and reducing its performance. With more nodes and flows in the network, XCP-b behavior is very unstable not coping correctly with the higher number of losses (Figure 3.20).

The Figures also show that TCP clearly outperforms the results obtained by XCP and RCP. The pure AIMD process of TCP allows it, when the network conditions are severe,

3.4 Performance Evaluation Results 0 500 1000 1500 2000 2500 3000 3500 4000 4 6 8 10 12 14 16 Number of Packets

Number of Mesh Nodes TCP Avg. Recv. Packets

XCP Avg. Recv. Packets XCP-b Avg. Recv. PacketsRCP Avg. Recv. Packets

Figure 3.17: Rate Based Protocols Average Average Received Packets - Variable Number of Mesh Nodes, 7 Mobile Nodes.

0 200 400 600 800 1000 1200 1400 1600 1800 2000 0 20 40 60 80 100 120 140 Throughput(Kbps)

Number of Simultaneous Flows TCP Avg. Throughput

XCP Avg. Throughput XCP-b Avg. ThroughputRCP Avg. Throughput

Figure 3.18: Rate Based Protocols Average Average Throughput - Ad-Hoc Scenario.

to maintain its stable behavior as it does not rely in information that must be updated in real-time, such as link capacity and available bandwidth. In a high utilized network with a high load of messages in transit, XCP and RCP are very unstable as they are not able to correctly infer the rate. This makes the rate based congestion control mechanism of XCP and RCP very inefficient and less fair than TCP. This behavior is also shown by Figure 3.18, Figure 3.19 and Figure 3.20, where it is possible to observe similar throughput values as TCP, with less received packets and worse delay vales.

10 100 1000 10000 0 20 40 60 80 100 120 140 Delay(ms)

Number of Simultaneous Flows TCP Avg. Delay

XCP Avg. Delay XCP-b Avg. DelayXCP Avg. Delay

Figure 3.19: Rate Based Protocols Average Average Delay - Ad-Hoc Scenario.

4000 6000 8000 10000 12000 14000 16000 18000 20000 22000 0 20 40 60 80 100 120 140 Number of Packets

Number of Simultaneous Flows

XCP Avg. Recv. Packets XCP-b Avg. Recv. Packets RCP Avg. Recv. Packets TCP Avg. Recv. Packets

Figure 3.20: Rate Based Protocols Average Average Received Packets - Ad-Hoc Scenario.